Anthropic has launched Claude Fable 5 for public access and introduced Claude Mythos 5 to select partners under Project Glasswing.
Anthropic has announced the public launch of its AI model, Claude Fable 5, while also introducing Claude Mythos 5 to a limited group of partners as part of Project Glasswing, a collaboration with the U.S. government.
While both models share the same foundational architecture, Claude Mythos 5 is notable for lacking some standard safety protocols. This new model serves as an upgrade to the previous Claude Mythos Preview, which was accessible in April to select cybersecurity experts and critical software infrastructure providers.
Looking ahead, Anthropic intends to extend access to Mythos 5 to individual researchers in biology and those involved in a trusted access program.
Claude Fable 5 is available to all users through the Claude API and consumption-based Enterprise plans. Users with Pro, Max, Team, and seat-based Enterprise subscriptions can utilize it without additional costs until June 22, 2026. After this date, credits will be necessary unless the company decides to extend the grace period.
This model features a mechanism that directs queries related to cybersecurity, biology, chemistry, and model distillation to the more advanced Claude Opus 4.8, which Anthropic claims happens in under 5% of sessions on average.
The pricing for both models is set at $10 per million input tokens and $50 per million output tokens. In May, Anthropic also released Claude Opus 4.8 and introduced a dynamic workflows feature for Claude Code.
